Reviewer's comment v1

Hello Yonas, this is Gerardo Flores 👋

You can find me on Discord as gerardo.flores.ds

I'm happy to review your project today 🙌

You can find my comments under the heading «Review». I will categorize my comments in green, blue or red boxes like this:

Success: if everything is done successfully

Remarks: if I can give some recommendations or ways to improve the project

Needs fixing: if the block requires some corrections. Work can't be accepted with the red comments

Please don't remove my comments. If you have any questions, don't hesitate to respond to my comments in a different section.

Student comments: For example like this

Reviewer's comment v1

General Feedback

You’ve submitted another project—great work! Your commitment to pushing through the challenges of this program is admirable.

After reviewing your submission, I consider it should be approved. You can find my more detailed feedback within your project file in the Reviewer's comment v1 section.

And of course, if you have any questions along the way, remember that you can always reach out to your tutor for any clarification.

Video Game Sales Analysis Project

Project Overview

In this project, you'll analyze video game sales data to identify patterns that determine a game's success. Working as an analyst for the online store Ice, you'll use this information to help plan future advertising campaigns.

Environment Setup and Required Libraries

Reviewer's comment v1

Great job adding additional information about the project goal. It would also be helpful to share the steps you plan to complete.

# Import all required libraries

import pandas as pd
import numpy as np
import math as mt
import plotly.express as px
from scipy import stats as st

Step 1: Loading and Initial Data Exploration

First, let's load our dataset and examine its basic properties:

# Load the dataset

data = pd.read_csv('/datasets/games.csv')

# Display basic information about the dataset

display(data)
print()
data.info()
print()
data.describe()

	Name	Platform	Year_of_Release	Genre	NA_sales	EU_sales	JP_sales	Other_sales	Critic_Score	User_Score	Rating
0	Wii Sports	Wii	2006.0	Sports	41.36	28.96	3.77	8.45	76.0	8	E
1	Super Mario Bros.	NES	1985.0	Platform	29.08	3.58	6.81	0.77	NaN	NaN	NaN
2	Mario Kart Wii	Wii	2008.0	Racing	15.68	12.76	3.79	3.29	82.0	8.3	E
3	Wii Sports Resort	Wii	2009.0	Sports	15.61	10.93	3.28	2.95	80.0	8	E
4	Pokemon Red/Pokemon Blue	GB	1996.0	Role-Playing	11.27	8.89	10.22	1.00	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...
16710	Samurai Warriors: Sanada Maru	PS3	2016.0	Action	0.00	0.00	0.01	0.00	NaN	NaN	NaN
16711	LMA Manager 2007	X360	2006.0	Sports	0.00	0.01	0.00	0.00	NaN	NaN	NaN
16712	Haitaka no Psychedelica	PSV	2016.0	Adventure	0.00	0.00	0.01	0.00	NaN	NaN	NaN
16713	Spirits & Spells	GBA	2003.0	Platform	0.01	0.00	0.00	0.00	NaN	NaN	NaN
16714	Winning Post 8 2016	PSV	2016.0	Simulation	0.00	0.00	0.01	0.00	NaN	NaN	NaN

16715 rows × 11 columns

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 16715 entries, 0 to 16714
Data columns (total 11 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   Name             16713 non-null  object 
 1   Platform         16715 non-null  object 
 2   Year_of_Release  16446 non-null  float64
 3   Genre            16713 non-null  object 
 4   NA_sales         16715 non-null  float64
 5   EU_sales         16715 non-null  float64
 6   JP_sales         16715 non-null  float64
 7   Other_sales      16715 non-null  float64
 8   Critic_Score     8137 non-null   float64
 9   User_Score       10014 non-null  object 
 10  Rating           9949 non-null   object 
dtypes: float64(6), object(5)
memory usage: 1.4+ MB

	Year_of_Release	NA_sales	EU_sales	JP_sales	Other_sales	Critic_Score
count	16446.000000	16715.000000	16715.000000	16715.000000	16715.000000	8137.000000
mean	2006.484616	0.263377	0.145060	0.077617	0.047342	68.967679
std	5.877050	0.813604	0.503339	0.308853	0.186731	13.938165
min	1980.000000	0.000000	0.000000	0.000000	0.000000	13.000000
25%	2003.000000	0.000000	0.000000	0.000000	0.000000	60.000000
50%	2007.000000	0.080000	0.020000	0.000000	0.010000	71.000000
75%	2010.000000	0.240000	0.110000	0.040000	0.030000	79.000000
max	2016.000000	41.360000	28.960000	10.220000	10.570000	98.000000

Since year_of_release is only missing 269 entries (1.6% of entries), and the year is an important metric in analysis, I will be substituting the median for missing values under said column. When it comes time for score and rating analysis I will be omitting entries that have missing values in this column, a significant number of them are missing, and substituting may skew our results.

# Check for duplicate entries

data.duplicated().sum()

#display(data[data['User_Score'] == 'tbd'])

0

Key Questions to Answer:

• What's the total number of records in our dataset?
• What data types are present in each column?
• Are there any obvious issues with the data?
• Do we see any immediate patterns or anomalies?

We have 16,715 entries, with 11 columns. Mostly float64 dtype, with the expected object type for Name, Platform, Genre, and Rating. We also have 'tbd' lurking within the user_score column, causing it to become object dtype.

Reviewer's comment v1

Well done!

Step 2: Data Preparation

2.1 Standardizing Column Names

# Convert column names to lowercase

data.columns = data.columns.str.lower()

# Verify the changes

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 16715 entries, 0 to 16714
Data columns (total 11 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   name             16713 non-null  object 
 1   platform         16715 non-null  object 
 2   year_of_release  16446 non-null  float64
 3   genre            16713 non-null  object 
 4   na_sales         16715 non-null  float64
 5   eu_sales         16715 non-null  float64
 6   jp_sales         16715 non-null  float64
 7   other_sales      16715 non-null  float64
 8   critic_score     8137 non-null   float64
 9   user_score       10014 non-null  object 
 10  rating           9949 non-null   object 
dtypes: float64(6), object(5)
memory usage: 1.4+ MB

column names successfully converted to lowercase

2.2 Data Type Conversion

# Check current data types

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 16715 entries, 0 to 16714
Data columns (total 11 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   name             16713 non-null  object 
 1   platform         16715 non-null  object 
 2   year_of_release  16446 non-null  float64
 3   genre            16713 non-null  object 
 4   na_sales         16715 non-null  float64
 5   eu_sales         16715 non-null  float64
 6   jp_sales         16715 non-null  float64
 7   other_sales      16715 non-null  float64
 8   critic_score     8137 non-null   float64
 9   user_score       10014 non-null  object 
 10  rating           9949 non-null   object 
dtypes: float64(6), object(5)
memory usage: 1.4+ MB

# Make changes to data types if necessary
# Describe the columns where the data types have been changed and why.

data['year_of_release'] = data['year_of_release'].fillna(2007)
data['year_of_release'] = data['year_of_release'].astype(int)
data.info()

#display(min(data['year_of_release']), max(data['year_of_release']))

#display(data)

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 16715 entries, 0 to 16714
Data columns (total 11 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   name             16713 non-null  object 
 1   platform         16715 non-null  object 
 2   year_of_release  16715 non-null  int64  
 3   genre            16713 non-null  object 
 4   na_sales         16715 non-null  float64
 5   eu_sales         16715 non-null  float64
 6   jp_sales         16715 non-null  float64
 7   other_sales      16715 non-null  float64
 8   critic_score     8137 non-null   float64
 9   user_score       10014 non-null  object 
 10  rating           9949 non-null   object 
dtypes: float64(5), int64(1), object(5)
memory usage: 1.4+ MB

I've decided to fill in the missing year values with the median, in order to convert the column to int type since we're only looking at the year, and not month/day. With the range being from 1980 - 2016

# Pay attention to the abbreviation TBD (to be determined). Specify how you intend to handle such cases.

data['user_score'] = data['user_score'].replace('tbd', np.nan)
data['user_score'] = data['user_score'].astype(float)
data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 16715 entries, 0 to 16714
Data columns (total 11 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   name             16713 non-null  object 
 1   platform         16715 non-null  object 
 2   year_of_release  16715 non-null  int64  
 3   genre            16713 non-null  object 
 4   na_sales         16715 non-null  float64
 5   eu_sales         16715 non-null  float64
 6   jp_sales         16715 non-null  float64
 7   other_sales      16715 non-null  float64
 8   critic_score     8137 non-null   float64
 9   user_score       7590 non-null   float64
 10  rating           9949 non-null   object 
dtypes: float64(6), int64(1), object(4)
memory usage: 1.4+ MB

The 'tbd' placeholder was preventing 'user_score' from being the correct dtype needed for future calculations. By changing these values to 'NaN', we can now make the proper dtype conversion, which will allow non-null entries to be used for calculations.

2.3 Handling Missing Values

# Examine missing values

display(data[data['name'].isna()])
print()
#display(data[data['user_score'].isna() & data['critic_score'].isna() & data['rating'].isna()])

	name	platform	year_of_release	genre	na_sales	eu_sales	jp_sales	other_sales	critic_score	user_score	rating
659	NaN	GEN	1993	NaN	1.78	0.53	0.00	0.08	NaN	NaN	NaN
14244	NaN	GEN	1993	NaN	0.00	0.00	0.03	0.00	NaN	NaN	NaN

We have many missing values within the critic_score, user_score, and rating columns. With only 2 entries who don't have a name or genre.

# Calculate percentage of missing values

ng_missing = (len(data[data['name'].isna()]) / len(data)) * 100
critic_missing = (len(data[data['critic_score'].isna()]) / len(data)) * 100
user_missing = (len(data[data['user_score'].isna()]) / len(data)) * 100
rating_missing = (len(data[data['rating'].isna()]) / len(data)) * 100

print(f'Percentage of entries missing "name" & "genre": {ng_missing}%')
print(f'Percentage of entries missing "critic_score": {critic_missing}%')
print(f'Percentage of entries missing "user_score": {user_missing}%')
print(f'Percentage of entries missing "rating": {rating_missing}%')

Percentage of entries missing "name" & "genre": 0.011965300628178284%
Percentage of entries missing "critic_score": 51.31917439425665%
Percentage of entries missing "user_score": 54.59168411606342%
Percentage of entries missing "rating": 40.478612025127134%

# Analyze patterns in missing values

#display(data[data['year_of_release'] < 1990])
#display(data[(data['year_of_release'] < 2000) & (data['platform'] == 'PS')])
display(data[data['year_of_release'] > 2010])

	name	platform	year_of_release	genre	na_sales	eu_sales	jp_sales	other_sales	critic_score	user_score	rating
16	Grand Theft Auto V	PS3	2013	Action	7.02	9.09	0.98	3.96	97.0	8.2	M
23	Grand Theft Auto V	X360	2013	Action	9.66	5.14	0.06	1.41	97.0	8.1	M
29	Call of Duty: Modern Warfare 3	X360	2011	Shooter	9.04	4.24	0.13	1.32	88.0	3.4	M
31	Call of Duty: Black Ops 3	PS4	2015	Shooter	6.03	5.86	0.36	2.38	NaN	NaN	NaN
33	Pokemon X/Pokemon Y	3DS	2013	Role-Playing	5.28	4.19	4.35	0.78	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...
16703	Strawberry Nauts	PSV	2016	Adventure	0.00	0.00	0.01	0.00	NaN	NaN	NaN
16707	Aiyoku no Eustia	PSV	2014	Misc	0.00	0.00	0.01	0.00	NaN	NaN	NaN
16710	Samurai Warriors: Sanada Maru	PS3	2016	Action	0.00	0.00	0.01	0.00	NaN	NaN	NaN
16712	Haitaka no Psychedelica	PSV	2016	Adventure	0.00	0.00	0.01	0.00	NaN	NaN	NaN
16714	Winning Post 8 2016	PSV	2016	Simulation	0.00	0.00	0.01	0.00	NaN	NaN	NaN

4022 rows × 11 columns

After running multiple parameters, I cannot seem to find a pattern in the entries with missing values. My initial guess was maybe the rating system was developed at a certain time period, perhaps pre 2000, but there exist entries with a non-null user or critic score pre 2000, even as far back as the 1980's.

Another theory I had was that maybe games only get ratings once per title, regardless of if that title released on multiple platforms (i.e. GTA V being released on X360 & PS3), but there exist entries that prove this theory false.

# Handle missing values based on analysis
# Your code here to handle missing values according to your strategy

data.dropna(subset=['name'], inplace=True)

data.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 16713 entries, 0 to 16714
Data columns (total 11 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   name             16713 non-null  object 
 1   platform         16713 non-null  object 
 2   year_of_release  16713 non-null  int64  
 3   genre            16713 non-null  object 
 4   na_sales         16713 non-null  float64
 5   eu_sales         16713 non-null  float64
 6   jp_sales         16713 non-null  float64
 7   other_sales      16713 non-null  float64
 8   critic_score     8137 non-null   float64
 9   user_score       7590 non-null   float64
 10  rating           9949 non-null   object 
dtypes: float64(6), int64(1), object(4)
memory usage: 1.5+ MB

I've decided to remove the 2 entries missing 'name' & 'platform', since their sales were so small they won't affect our analysis. These 2 entries account for .01% of our entire data set.

Earlier in this analysis I substituted the missing values under 'year_of_release' with the median (2007), I feel confident in this considering the mean is (2006.48)

# Why do you think the values are missing? Give possible reasons.
# Explain why you filled in the missing values as you did or why you decided to leave them blank.

Percentage of entries missing "critic_score": 51.31917439425665%

Percentage of entries missing "user_score": 54.59168411606342%

Percentage of entries missing "rating": 40.478612025127134%

With the quantity of games here, and the subjectivity of scores, it's possible that a score could either not be settled upon when gathering this data (which would explain the 1985 Super Mario Bro's), or some of these titles didn't sell enough to be considered for a score. The reality is likely a combination of both, but with the sheer amount of data missing, I believe the best course of action would be to leave these blank, and create new dataframes containing the valid entries when it comes time to analyze the scores and rating.

Reviewer's comment v1

Well done!

2.4 Calculate Total Sales

# Calculate total sales across all regions and put them in a different column

data['total_sales'] = data['na_sales'] + data['eu_sales'] + data['jp_sales'] + data['other_sales']

data.info()
display(data.sample(30, random_state=999))

<class 'pandas.core.frame.DataFrame'>
Int64Index: 16713 entries, 0 to 16714
Data columns (total 12 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   name             16713 non-null  object 
 1   platform         16713 non-null  object 
 2   year_of_release  16713 non-null  int64  
 3   genre            16713 non-null  object 
 4   na_sales         16713 non-null  float64
 5   eu_sales         16713 non-null  float64
 6   jp_sales         16713 non-null  float64
 7   other_sales      16713 non-null  float64
 8   critic_score     8137 non-null   float64
 9   user_score       7590 non-null   float64
 10  rating           9949 non-null   object 
 11  total_sales      16713 non-null  float64
dtypes: float64(7), int64(1), object(4)
memory usage: 1.7+ MB

	name	platform	year_of_release	genre	na_sales	eu_sales	jp_sales	other_sales	critic_score	user_score	rating	total_sales
3256	Prince of Persia: The Forgotten Sands	X360	2010	Action	0.30	0.26	0.00	0.06	74.0	7.8	T	0.62
10572	Football Director DS	DS	2008	Sports	0.00	0.10	0.00	0.00	NaN	NaN	NaN	0.10
9920	Spider-Man: Edge of Time	3DS	2011	Action	0.08	0.03	0.00	0.01	50.0	6.4	T	0.12
10204	Snowboard Kids 2	N64	1999	Sports	0.09	0.02	0.00	0.00	NaN	NaN	NaN	0.11
13602	Record of Agarest War: Marriage	PSP	2012	Role-Playing	0.00	0.00	0.04	0.00	NaN	NaN	NaN	0.04
3156	Real Heroes: Firefighter	Wii	2009	Action	0.56	0.04	0.00	0.05	67.0	8.2	E10+	0.65
15800	Canada Hunt	Wii	2010	Sports	0.02	0.00	0.00	0.00	NaN	NaN	T	0.02
4762	Tak and the Power of Juju	GC	2003	Platform	0.31	0.08	0.00	0.01	71.0	8.2	E	0.40
8898	High Velocity Bowling	PS3	2010	Sports	0.13	0.00	0.00	0.01	66.0	8.1	E10+	0.14
13190	Harukanaru Toki no Naka de 3	PS2	2004	Adventure	0.00	0.00	0.05	0.00	NaN	NaN	NaN	0.05
5243	Formula 1 Championship Edition	PS	1997	Racing	0.06	0.04	0.23	0.02	NaN	NaN	NaN	0.35
16326	Victory Spike	PS	1996	Sports	0.00	0.00	0.01	0.00	NaN	NaN	NaN	0.01
5991	Imagine: Fashion Designer World Tour	DS	2009	Simulation	0.27	0.00	0.00	0.02	NaN	NaN	E	0.29
8472	TransFormers Animated: The Game	DS	2008	Action	0.15	0.00	0.00	0.01	69.0	NaN	E	0.16
11997	The Legend of Heroes VII: The Trail of Zero	PSV	2012	Role-Playing	0.00	0.00	0.07	0.00	NaN	NaN	NaN	0.07
9040	Thor: God of Thunder	X360	2011	Action	0.09	0.04	0.00	0.01	38.0	5.7	T	0.14
3672	Top Gun: Fire at Will!	PS	1996	Misc	0.30	0.21	0.00	0.04	NaN	NaN	NaN	0.55
13876	Doraemon: Shin Nobita no Nihon Tanjou	3DS	2016	Misc	0.00	0.00	0.04	0.00	NaN	NaN	NaN	0.04
5080	Imagine: Party Babyz	Wii	2008	Simulation	0.33	0.02	0.00	0.03	NaN	6.1	E	0.38
11266	Monster Jam: Urban Assault	PSP	2008	Racing	0.08	0.00	0.00	0.01	NaN	NaN	E	0.09
10364	Pro Yakyuu Team o Tsukurou! 2	DS	2009	Sports	0.00	0.00	0.11	0.00	NaN	NaN	NaN	0.11
3825	Tiger Woods PGA Tour 11	X360	2010	Sports	0.28	0.19	0.00	0.05	79.0	7.8	E	0.52
9462	Mega Man Anniversary Collection	XB	2005	Platform	0.10	0.03	0.00	0.00	80.0	9.2	E	0.13
9336	Rogue Warrior	PS3	2009	Shooter	0.11	0.01	0.00	0.01	27.0	2.8	M	0.13
2802	Practise English!	DS	2007	Misc	0.00	0.00	0.73	0.00	NaN	NaN	NaN	0.73
3984	Star Wars: Clone Wars	PS2	2002	Shooter	0.24	0.19	0.00	0.06	NaN	NaN	NaN	0.49
2936	Minecraft: Story Mode	PS4	2015	Adventure	0.25	0.33	0.00	0.11	NaN	NaN	NaN	0.69
722	Spyro: Season of Ice	GBA	2001	Platform	1.29	0.83	0.00	0.11	74.0	8.0	E	2.23
12836	Nobunaga's Ambition: Sphere of Influence - Sen...	PS4	2016	Misc	0.01	0.00	0.05	0.00	NaN	NaN	NaN	0.06
7160	NCAA Football 2005	GC	2004	Sports	0.17	0.04	0.00	0.01	88.0	9.0	E	0.22

Step 3: Analyzing Video Game Sales Data

3.1 Temporal Analysis of Game Releases

Let's first examine the distribution of game releases across different years to understand our data's coverage and significance:

# Create a DataFrame with game releases by year

year_data = data[['name', 'platform', 'year_of_release']]
year_data = year_data.sort_values(by='year_of_release')

display(year_data)
year_data.info()

	name	platform	year_of_release
1968	Defender	2600	1980
6300	Bridge	2600	1980
1764	Kaboom!	2600	1980
546	Missile Command	2600	1980
5360	Freeway	2600	1980
...	...	...	...
14404	Prison Architect	PS4	2016
11168	Terraria	3DS	2016
14414	Atelier Shallie Plus: Alchemists of the Dusk Sea	PSV	2016
11243	Assassin's Creed Chronicles	PS4	2016
16714	Winning Post 8 2016	PSV	2016

16713 rows × 3 columns

<class 'pandas.core.frame.DataFrame'>
Int64Index: 16713 entries, 1968 to 16714
Data columns (total 3 columns):
 #   Column           Non-Null Count  Dtype 
---  ------           --------------  ----- 
 0   name             16713 non-null  object
 1   platform         16713 non-null  object
 2   year_of_release  16713 non-null  int64 
dtypes: int64(1), object(2)
memory usage: 522.3+ KB

# Visualize the distribution of games across years

fig = px.histogram(year_data, x='year_of_release', width=1000, height=600)

fig.update_layout(title_text='Quantity of Games Released by Year',
                  xaxis_title_text='Year', yaxis_title_text='Quantity')
fig.show()

# Display summary statistics for each year

#Summary Statistics Calculation for the Entire Sample:

def mean_var_stdv(series):
    mean = series.mean()
    var = series.var()
    stdv = mt.sqrt(var)
    return f'Mean: {mean}\nVariance: {var}\nSt.Deviation: {stdv}'

print(f'Temporal Summary Statistics for all Observable Years:\n\n{mean_var_stdv(year_data["year_of_release"])}\n')
print(f'-~-~-~-' * 21)
print(f'\nDataFrame Containing Amount of Games Released and Corresponding Percentage, Grouped by Platform and Year:\n')

#Creating new DataFrame containing the amount of titles released each year by platform, as well as the corresponding share percentage (ex. 10 releases in one year / 100 releases across all years = 10% of sales in that year)

total_by_year = year_data.groupby(['platform', 'year_of_release'])['name'].count().reset_index()
total_by_year = total_by_year.rename(columns={'year_of_release' : 'year', 'name' : 'titles_released_count'})

lifetime_total_series = total_by_year.groupby('platform')['titles_released_count'].transform(sum)
total_by_year['platform_lifetime_percentage'] = ((total_by_year['titles_released_count'] / lifetime_total_series) * 100).map('{:.2f}%'.format)

display(total_by_year)

#display(lifetime_total_series)

Temporal Summary Statistics for all Observable Years:

Mean: 2006.4945252198886
Variance: 33.970305711920766
St.Deviation: 5.828405074453968

-~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~--~-~-~-

DataFrame Containing Amount of Games Released and Corresponding Percentage, Grouped by Platform and Year:

	platform	year	titles_released_count	platform_lifetime_percentage
0	2600	1980	9	6.77%
1	2600	1981	46	34.59%
2	2600	1982	36	27.07%
3	2600	1983	11	8.27%
4	2600	1984	1	0.75%
...	...	...	...	...
239	XB	2008	1	0.12%
240	XOne	2013	19	7.69%
241	XOne	2014	61	24.70%
242	XOne	2015	80	32.39%
243	XOne	2016	87	35.22%

244 rows × 4 columns

1. Temporal Summary Statistics match what we can see in the previously shown histogram, with the peak lying somewhere between 2006 - 2010. There is a higher volume of data on the left side of our median, causing the mean to be slightly lower than if we had a balanced distribution. This is also proven by the fact that our (-3 sigma)(2006.49 - (5.83*3)) or (1989) is observable, but only our (+1 sigma)(2006.49 + 5.83) or (2012) is visible.

2. Summary Statistics by Platform and Year, this is a DataFrame containing every observable year in our dataset, with the amount of games released on each platform serving as entries. The percentage of each entry is calculated using the total games released per platform per year, divided by the total number of games released over the lifespan of each platform.

Questions to Consider:

• Which years show significant numbers of game releases?
• Are there any notable trends or patterns in the number of releases?
• Is there enough recent data to make predictions for 2017?

Our first sharp increase in games released was in the year 2000, seeing a 236.86% increase in the next 2 years (2000 - 2002). Something to consider: "The Matrix" trilogy was in the process of unfolding (1999 - 2003), and with it brought a wider audience into the 'cyberspace', likely contributing to this surge. A bulk of our data lies between the years of 2007 - 2011, seeing a consistent increase in the gaming market. With the introduction of YouTube in 2005, gaming would see a massive market shift, partly explaining the huge increase mentioned previously.

I believe with our data set it is possible to make predictions based on observable trends, however, there is no certainty in the future. Weather predictions are a great example of this.

Reviewer's comment v1

Great job!

3.2 Platform Sales Analysis Over Time

Now let's analyze how sales vary across platforms and years:

# Calculate total sales by platform and year

plt_yr_total = data.groupby(['platform', 'year_of_release'])['total_sales'].sum()
plt_yr_total = plt_yr_total.unstack(level='year_of_release')
plt_yr_total = plt_yr_total.fillna(0)

display(plt_yr_total)
plt_yr_total.info()

#display(data['platform'].value_counts())

#year_plt_sales = data[['name', 'platform', 'total_sales', 'year_of_release']]
#year_plt_sales = year_plt_sales.sort_values(by='year_of_release')
#year_plt_sales.info()
#display(year_plt_sales)

year_of_release	1980	1981	1982	1983	1984	1985	1986	1987	1988	1989	...	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016
platform
2600	11.38	35.68	28.88	5.84	0.27	0.45	0.67	1.94	0.74	0.63	...	10.50	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
3DO	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
3DS	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	1.19	0.00	0.00	0.00	63.20	51.36	56.57	43.76	27.78	15.14
DC	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.02	0.04	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
DS	0.00	0.00	0.00	0.00	0.00	0.02	0.00	0.00	0.00	0.00	...	150.28	145.31	119.54	85.02	26.18	11.01	1.54	0.00	0.00	0.00
GB	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	1.43	64.97	...	1.03	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
GBA	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	8.37	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
GC	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	2.47	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
GEN	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
GG	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
N64	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.67	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NES	0.00	0.00	0.00	10.96	50.08	53.44	36.41	19.76	45.01	7.85	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NG	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
PC	0.00	0.00	0.00	0.00	0.00	0.04	0.00	0.00	0.03	0.00	...	13.04	12.42	16.91	24.28	35.03	23.22	12.38	13.28	8.52	5.25
PCFX	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
PS	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	3.28	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
PS2	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	98.20	53.90	26.40	5.64	0.45	0.00	0.00	0.00	0.00	0.00
PS3	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	81.50	118.52	130.93	142.17	156.78	107.36	113.25	47.76	16.82	3.60
PS4	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	25.99	100.00	118.90	69.25
PSP	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	51.45	34.55	37.78	35.04	17.82	7.69	3.14	0.24	0.12	0.00
PSV	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.26	0.00	0.00	0.00	4.63	16.19	10.59	11.90	6.25	4.25
SAT	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
SCD	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
SNES	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
TG16	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
WS	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
Wii	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	169.10	171.32	206.97	127.95	59.65	21.71	8.59	3.75	1.14	0.18
WiiU	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	17.56	21.65	22.03	16.35	4.60
X360	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	105.59	135.26	120.29	170.03	143.84	99.74	88.58	34.74	11.96	1.52
XB	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	6.72	0.18	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
XOne	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	...	0.00	0.00	0.00	0.00	0.00	0.00	18.96	54.07	60.14	26.15

31 rows × 37 columns

<class 'pandas.core.frame.DataFrame'>
Index: 31 entries, 2600 to XOne
Data columns (total 37 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   1980    31 non-null     float64
 1   1981    31 non-null     float64
 2   1982    31 non-null     float64
 3   1983    31 non-null     float64
 4   1984    31 non-null     float64
 5   1985    31 non-null     float64
 6   1986    31 non-null     float64
 7   1987    31 non-null     float64
 8   1988    31 non-null     float64
 9   1989    31 non-null     float64
 10  1990    31 non-null     float64
 11  1991    31 non-null     float64
 12  1992    31 non-null     float64
 13  1993    31 non-null     float64
 14  1994    31 non-null     float64
 15  1995    31 non-null     float64
 16  1996    31 non-null     float64
 17  1997    31 non-null     float64
 18  1998    31 non-null     float64
 19  1999    31 non-null     float64
 20  2000    31 non-null     float64
 21  2001    31 non-null     float64
 22  2002    31 non-null     float64
 23  2003    31 non-null     float64
 24  2004    31 non-null     float64
 25  2005    31 non-null     float64
 26  2006    31 non-null     float64
 27  2007    31 non-null     float64
 28  2008    31 non-null     float64
 29  2009    31 non-null     float64
 30  2010    31 non-null     float64
 31  2011    31 non-null     float64
 32  2012    31 non-null     float64
 33  2013    31 non-null     float64
 34  2014    31 non-null     float64
 35  2015    31 non-null     float64
 36  2016    31 non-null     float64
dtypes: float64(37)
memory usage: 9.2+ KB

# Create a heatmap of platform sales over time

fig = px.imshow(plt_yr_total, labels=dict(x="Year", y="Platform", color="Total Sales"),
                x=plt_yr_total.columns, y=plt_yr_total.index, width=1000, height=800, 
                title='Total Game Sales (USD Million) by Platform and Year')
fig.show()

# Identify platforms with declining sales

Any platform that has been usurped by newer technology will begin to decline, as we can see with the platforms 'PS, PS2, PS3, PS4'. When sales peak and begin to decline, that is when the market seemingly migrates to the next big thing. There are additional "unknown" factors that contribute to sales, example being the announcement of a newer console may influence people to halt consumption and wait for the newer one. Another factor worth consideration is console exclusive titles that may switch certain consumers from one platform to the other, if they do not wish to possess multiple.

Questions to Consider:

• Which platforms show consistent sales over time?
• Can you identify platforms that have disappeared from the market?
• What's the typical lifecycle of a gaming platform?

The leading platform in sales consistency has to be 'PC' at 31 years (1985 - 2016), which makes sense since there are no definitive hardware constraints by which to categorize PC from generation to generation.

'GB' comes in second, with consistent sales accross a surprising 13 year span (1988 - 2001). Finally, 'NES' comes in third for consistency in sales over time with 11 years (1983 -1994).

When looking for 'consistency', we want the most stable record over a given time period. Not including platforms that saw meteoric rises before succumbing to their successors (in the case of Playstation's four generations, and multiple handhelds).

Reviewer's comment v1

Great job!

3.3 Determining Relevant Time Period

Based on your analysis above, determine the appropriate time period for predicting 2017 sales:

# Your code here to filter the dataset to relevant years
# Example:
# relevant_years = [XXXX, XXXX, XXXX] # Replace with your chosen years
# df_relevant = df[df['year_of_release'].isin(relevant_years)]

# Justify your choice with data

select_years = [2011, 2012, 2013, 2014, 2015, 2016]
data_select_years = data[['platform', 'year_of_release', 'total_sales']]
data_select_years = data_select_years[data_select_years['year_of_release'].isin(select_years)].sort_values(by='year_of_release')
data_select_years = data_select_years.groupby(['platform', 'year_of_release'])['total_sales'].sum().reset_index()

#data_select_years = data_select_years.unstack(level='year_of_release').fillna(0)

display(data_select_years.head(20))

#data_select_years.info()

	platform	year_of_release	total_sales
0	3DS	2011	63.20
1	3DS	2012	51.36
2	3DS	2013	56.57
3	3DS	2014	43.76
4	3DS	2015	27.78
5	3DS	2016	15.14
6	DS	2011	26.18
7	DS	2012	11.01
8	DS	2013	1.54
9	PC	2011	35.03
10	PC	2012	23.22
11	PC	2013	12.38
12	PC	2014	13.28
13	PC	2015	8.52
14	PC	2016	5.25
15	PS2	2011	0.45
16	PS3	2011	156.78
17	PS3	2012	107.36
18	PS3	2013	113.25
19	PS3	2014	47.76

Document Your Decision:

• What years did you select and why?
• How does this period reflect current market conditions?
• What factors influenced your decision?

I've decided to go with 2011 - 2016, since it should be easier to make a prediction with a stable trajectory. Kind of like analyzing the influence of gravity on a projectile.

As for current market conditions, we don't know for certain the reason of declining title releases over time. However, we have to acknowledge the trending decline, since titles have to be released in order for sales to occur, which is the primary reason for my time frame selection.

Reviewer's comment v1

Great time frame decision!

3.4 Platform Performance Analysis

Using your selected time period, let's analyze platform performance:

# Analyze platform sales trends

display(data_select_years.head(10))

	platform	year_of_release	total_sales
0	3DS	2011	63.20
1	3DS	2012	51.36
2	3DS	2013	56.57
3	3DS	2014	43.76
4	3DS	2015	27.78
5	3DS	2016	15.14
6	DS	2011	26.18
7	DS	2012	11.01
8	DS	2013	1.54
9	PC	2011	35.03

With the first 10 rows visible, we can see that the 'DS' was at the tail end of it's lifecycle (2011 - 2013), 'passing the torch' to the next generation of console '3DS'. Both platforms would see sales concurrently (2011 - 2013) until titles were no longer being developed for the 'parent' console, or 'DS', giving way to the 'child' console, or '3DS'. This trend occurs accross multiple platforms in this data analysis, the only exception being 'PC', where no generational gaps can be distinguished.

# Sort platforms by total sales

#data_select_years['total_sales'] = data_select_years.sum(axis=1)
data_select_years = data_select_years.sort_values(by='total_sales')

display(data_select_years.head(20))
data_select_years.info()

	platform	year_of_release	total_sales
30	PSP	2015	0.12
42	Wii	2016	0.18
29	PSP	2014	0.24
15	PS2	2011	0.45
41	Wii	2015	1.14
53	X360	2016	1.52
8	DS	2013	1.54
28	PSP	2013	3.14
21	PS3	2016	3.60
40	Wii	2014	3.75
36	PSV	2016	4.25
47	WiiU	2016	4.60
31	PSV	2011	4.63
14	PC	2016	5.25
35	PSV	2015	6.25
27	PSP	2012	7.69
13	PC	2015	8.52
39	Wii	2013	8.59
33	PSV	2013	10.59
7	DS	2012	11.01

<class 'pandas.core.frame.DataFrame'>
Int64Index: 58 entries, 30 to 16
Data columns (total 3 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   platform         58 non-null     object 
 1   year_of_release  58 non-null     int64  
 2   total_sales      58 non-null     float64
dtypes: float64(1), int64(1), object(1)
memory usage: 1.8+ KB

# Visualize top platforms
# Calculate year-over-year growth for each platform
# Your code here to calculate and visualize platform growth rates

#fig = px.histogram(data_select_years, x='platform', y='total_sales', 
                   #color='year_of_release', barmode='group', height=500, 
                   #labels={'total_sales' : 'Total Sales'})

#Visualizing platform performance by year, and changes in sales over time:

data_select_years = data_select_years.sort_values(by='year_of_release')

import plotly.express.colors as px_colors

fig = px.line(data_select_years, x='year_of_release', y='total_sales',
                 color='platform', symbol='platform', height=800,
                 labels={'total_sales' : 'Total Sales (USD Million)', 'year_of_release' : 'Year'},
              title='Which Platform Sold the Most Titles from 2011 to 2016?',
             color_discrete_sequence=px_colors.qualitative.Dark24)

fig.show()

I've chosen to use a line graph in order to showcase both the yearly standing, and changes over time for each console. With the amount of platforms on display here, it was vital to ensure visual clarity, by utilizing an external color pack and symbols.

As stated earlier, there are multiple generations of platform on display here. New hardware launches and there are a few years of simultaneous market activity in both, before the predeccessor is inevitably rendered obsolete when titles cease to launch. Some platforms have milder and more consistent market trends (Wii & WiiU) than the top selling platforms (PS3 & PS4), but loosely follow the same pattern of generational succession.

Reviewer's comment v1

Well done!

3.5 Sales Distribution Analysis

Let's examine the distribution of sales across platforms:

# Create box plot of sales by platform

fig = px.box(data_select_years, x='platform', y='total_sales',
            color='platform', height=700, labels={'total_sales' : 'Total Sales (USD Million)', 'platform' : 'Platform'},
            points='all', title='2011 to 2016: Annual Sales Distribution Analysis')
fig.show()

#display(data_select_years)

This box plot shows us a similar pattern, but in different format. With the older generations of platform beginning to trend downwards, the closer to 0 the min value is, the closer to the end of the platforms market life cycle. Sales are still reffering to games sold under each platform.

# Calculate detailed statistics for each platform

boxplot_summary_stats = data_select_years.groupby('platform')['total_sales'].describe()
display(boxplot_summary_stats)

	count	mean	std	min	25%	50%	75%	max
platform
3DS	6.0	42.968333	18.278261	15.14	31.7750	47.560	55.2675	63.20
DS	3.0	12.910000	12.429397	1.54	6.2750	11.010	18.5950	26.18
PC	6.0	16.280000	11.007403	5.25	9.4850	12.830	20.7350	35.03
PS2	1.0	0.450000	NaN	0.45	0.4500	0.450	0.4500	0.45
PS3	6.0	74.261667	60.695667	3.60	24.5550	77.560	111.7775	156.78
PS4	4.0	78.535000	40.567917	25.99	58.4350	84.625	104.7250	118.90
PSP	5.0	5.802000	7.387558	0.12	0.2400	3.140	7.6900	17.82
PSV	6.0	8.968333	4.729695	4.25	5.0350	8.420	11.5725	16.19
Wii	6.0	15.836667	22.864073	0.18	1.7925	6.170	18.4300	59.65
WiiU	5.0	16.438000	7.068244	4.60	16.3500	17.560	21.6500	22.03
X360	6.0	63.396667	56.074638	1.52	17.6550	61.660	96.9500	143.84
XOne	4.0	39.830000	20.313977	18.96	24.3525	40.110	55.5875	60.14

A more refined look at our data, but still the same message is conveyed: min values and their proximity to the mean tells us what kind of trajectory our platforms are on. 'PC' emerging in the 1990's and keeping with a consistent trend even up until 2016. On the other hand we have the 'PS3', the best performing console within this sample, but rapidly approaching zero, as the successor 'PS4' enters the market.

3.6 Review Score Impact Analysis

Select a popular platform and analyze how reviews affect sales:

# Choose a popular platform based on your previous analysis

popular_platform = data[(data['platform'] == 'PS3') & (data['year_of_release'] >= 2011)][['name', 'year_of_release', 'critic_score', 'user_score', 'total_sales']].dropna()

display(popular_platform)

	name	year_of_release	critic_score	user_score	total_sales
16	Grand Theft Auto V	2013	97.0	8.2	21.05
34	Call of Duty: Black Ops II	2012	83.0	5.3	13.79
37	Call of Duty: Modern Warfare 3	2011	88.0	3.2	13.33
69	Call of Duty: Ghosts	2013	71.0	2.6	9.36
81	FIFA Soccer 13	2012	88.0	6.6	8.17
...	...	...	...	...	...
15533	Battle Princess of Arcadias	2013	69.0	7.9	0.02
15567	Tom Clancy's Splinter Cell Classic Trilogy HD	2011	67.0	6.8	0.02
15742	Invizimals: The Lost Kingdom	2013	50.0	5.0	0.02
16146	Don Bradman Cricket 14	2014	73.0	6.9	0.01
16230	Madden NFL 13	2012	83.0	5.5	0.01

342 rows × 5 columns

The console that saw the most sales in our selective analysis was undoubtedly the 'PS3', with the largest mean/stdv pairing of all platforms. This lets us peer into the vastness of this platform and it's performative reach.

Since we're still under "3.x" I will continue the selected year trend to conduct further analysis until "4.0", or explicitly said otherwise.

As I mentioned at the beginning of this analysis, I will be omitting all entries with missing values in the event of 'critic_score' and 'user_score' analysis. As these entries have no value in this angle of observation.

# Create scatter plots for both critic and user scores

popular_platform_twinbox = popular_platform.melt(id_vars=['name', 'year_of_release', 'total_sales'],
                                                 value_vars=['critic_score', 'user_score'], var_name='score_type',
                                                 value_name='score')

popular_platform_twinbox['score'] = popular_platform_twinbox.apply(lambda row: row['score'] * 10 if row['score_type'] == 'user_score' else
                                                          row['score'], axis=1)

#display(popular_platform_twinbox)

fig = px.strip(popular_platform_twinbox, x='year_of_release', y='score',
                 color='score_type', facet_col='score_type', height=900, hover_data=['name'],
               labels={'year_of_release' : 'Year', 'score' : 'Score (out of 100)'},
              title='How Similar are the Review Scores of Critics and Users?')
fig.update_traces(jitter=1)
fig.show()

Scores between the two sources seem to be mildy similar at a glance, with some entries being within 5 points of each other (2013 'The Last of Us' at 95 vs 91), and some being over 10 points apart (2013 'Grand Theft Auto V' at 97 vs 82). There may be any number of reasons for the differing scores, but in my experience within this domain: critics tend to have different motives for playing vs users (duty vs leisure).

It's also worth considering that critics may receive the title for free before review, since their scores can influence the market. With this in mind, reviewing a game that costs nothing may positively skew the players experience (critic_score). While paying for the title sets an expectation of said product (user_score). This dynamic will not be true in every case, and could be one of many factors, but in the case of 'MindJack' released in 2011, this potentially explains the massive gap between a Critic Score of 44, and User Score of 17.

# Critic Scores

critic_scores = popular_platform['critic_score']

# User Scores

user_scores = popular_platform['user_score']

# Calculate correlations

score_correlation = critic_scores.corr(user_scores)

print(score_correlation)

0.5633145093196069

With a Correlation Coefficient of .56 between the two score types, we can say there is a moderate positive correlation. Critics and Users have similar taste, but not identical.

3.7 Cross-Platform Comparison

Compare sales performance of games across different platforms:

# Find games released on multiple platforms

cross_platform = data[(data['year_of_release'] >= 2011) & (data['total_sales'] >= 1.0)][['name', 'year_of_release', 'platform', 'total_sales']]
cross_platform = cross_platform[cross_platform['name'].duplicated(keep=False)]

display(cross_platform)

cross_platform.info()

	name	year_of_release	platform	total_sales
16	Grand Theft Auto V	2013	PS3	21.05
23	Grand Theft Auto V	2013	X360	16.27
29	Call of Duty: Modern Warfare 3	2011	X360	14.73
31	Call of Duty: Black Ops 3	2015	PS4	14.63
34	Call of Duty: Black Ops II	2012	PS3	13.79
...	...	...	...	...
2034	Skylanders SWAP Force	2013	PS3	1.02
2036	Sonic Generations	2011	3DS	1.01
2048	Need for Speed: Most Wanted	2012	PSV	1.01
2050	Assassin's Creed: Rogue	2014	X360	1.01
2091	Dark Souls	2011	X360	1.00

254 rows × 4 columns

<class 'pandas.core.frame.DataFrame'>
Int64Index: 254 entries, 16 to 2091
Data columns (total 4 columns):
 #   Column           Non-Null Count  Dtype  
---  ------           --------------  -----  
 0   name             254 non-null    object 
 1   year_of_release  254 non-null    int64  
 2   platform         254 non-null    object 
 3   total_sales      254 non-null    float64
dtypes: float64(1), int64(1), object(2)
memory usage: 9.9+ KB

# Compare sales across platforms for these games
# Your code here to analyze and visualize cross-platform performance

fig = px.scatter(cross_platform, x='name', y='total_sales',
                 color='name', symbol='platform', height=800, size='total_sales',
                 labels={'total_sales' : 'Total Sales (USD Million)', 'name' : 'Game Title', 'platform' : 'Platform'},
                 title='How Well do Cross-Platform Games Perform?')

fig.show()

This is a ton of data points, and for the sake of visual clarity I've decided to ommit any games that did not cross 1 million usd in sales.

With that in mind, we can see that games releasing on multiple platforms tend to net more sales overall. With some of the largest titles releasing on consecutive generations of platform, which we can easily see thanks to plotly's wonderful legend layout.

Earlier platform popularity observations extend to this analysis, seeing that there is a heirarchy of which console sells the most games. 'PS3' will almost always lead, with 'X360' following close behind, and the next generations ('PS4' & 'XOne') making their debut, with some titles re-releasing on the newer hardware.

3.8 Genre Analysis

Finally, let's examine the distribution of games by genre:

# Analyze genre performance

genre_analysis = data[data['year_of_release'] >= 2011][['genre', 'total_sales']]
genre_analysis = genre_analysis.groupby('genre')['total_sales'].sum().reset_index()

display(genre_analysis)

genre_analysis.info()

	genre	total_sales
0	Action	557.80
1	Adventure	45.04
2	Fighting	66.90
3	Misc	139.42
4	Platform	88.68
5	Puzzle	9.85
6	Racing	88.31
7	Role-Playing	245.63
8	Shooter	402.91
9	Simulation	50.26
10	Sports	237.15
11	Strategy	22.16

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 12 entries, 0 to 11
Data columns (total 2 columns):
 #   Column       Non-Null Count  Dtype  
---  ------       --------------  -----  
 0   genre        12 non-null     object 
 1   total_sales  12 non-null     float64
dtypes: float64(1), object(1)
memory usage: 320.0+ bytes

# Sort genres by total sales

genre_analysis = genre_analysis.sort_values(by='total_sales', ascending=False)

display(genre_analysis)

	genre	total_sales
0	Action	557.80
8	Shooter	402.91
7	Role-Playing	245.63
10	Sports	237.15
3	Misc	139.42
4	Platform	88.68
6	Racing	88.31
2	Fighting	66.90
9	Simulation	50.26
1	Adventure	45.04
11	Strategy	22.16
5	Puzzle	9.85

# Visualize genre distribution

fig = px.bar(genre_analysis, x='genre', y='total_sales', height=600,
             color='genre', labels={'total_sales' : 'Total Sales (USD Million)', 'genre' : 'Genre'},
             title='Which Genre Sells the Most?')

fig.update_layout(showlegend=False)
fig.show()

# Calculate market share for each genre

sales_sum = genre_analysis['total_sales'].sum()

genre_analysis['market_share'] = ((genre_analysis['total_sales'] / sales_sum) * 100).map('{:.2f}%'.format)

display(genre_analysis)

#market_share_100 = genre_analysis['market_share'].sum()
#print(market_share_100)

	genre	total_sales	market_share
0	Action	557.80	28.54%
8	Shooter	402.91	20.62%
7	Role-Playing	245.63	12.57%
10	Sports	237.15	12.14%
3	Misc	139.42	7.13%
4	Platform	88.68	4.54%
6	Racing	88.31	4.52%
2	Fighting	66.90	3.42%
9	Simulation	50.26	2.57%
1	Adventure	45.04	2.30%
11	Strategy	22.16	1.13%
5	Puzzle	9.85	0.50%

Action games are leading the pack substantially, with Shooter trailing close behind. These 2 genres make up almost half (~49%) of the total gaming market recently (2011 - 2016) by themselves. Again, this is what we have been seeing in previous observations. 'Grand Theft Auto V' and the 'Call of Duty' series are the best selling titles in this sample, and having been released on multiple platforms accross multiple generations secures their market positions further.

Reviewer's comment v1

Great job!

Key Questions for Genre Analysis:

• Which genres consistently perform well?
• Are there any genres showing recent growth or decline?
• How does the average performance vary across genres?

Step 4: Regional Market Analysis and User Profiles

In this section, we will analyze the gaming market characteristics across three major regions: North America (NA), Europe (EU), and Japan (JP). Our analysis will focus on platform preferences, genre popularity, and the impact of ESRB ratings in each region.

4.1 Regional Platform Analysis

Let's begin by examining platform performance across different regions:

# Function to analyze platform performance by region

def regional_sales(data: pd.DataFrame, desired_region: str) -> pd.DataFrame:
    
    selected = data[data['year_of_release'] >= 2011]
    columns = ['platform', desired_region]
    
    result = selected[columns].groupby('platform')[desired_region].sum().reset_index()
    return result.sort_values(by=desired_region, ascending=False)

#jp_user = data[(data['year_of_release'] >= 2011) & (data['jp_sales'] >= 1.0)][['name', 'platform', 'year_of_release',
                                                                               #'genre', 'jp_sales', 'critic_score',
                                                                               #'user_score', 'rating']].sort_values(by='jp_sales')
#display(jp_user)
#jp_user.info()

#data.info()
#display(data)

# Analyze each region

regional_plotly = regional_sales(data, 'na_sales').merge(regional_sales(data, 'eu_sales'), on='platform', how='outer')
regional_plotly = regional_plotly.merge(regional_sales(data, 'jp_sales'), on='platform', how='outer')
regional_plotly = regional_plotly.rename(columns={'platform' : 'Platform', 'na_sales' : 'NA Sales', 'eu_sales' : 'EU Sales', 'jp_sales' : 'JP Sales'})

display(regional_plotly)

#display(regional_sales(data, 'na_sales'), regional_sales(data, 'eu_sales'), regional_sales(data, 'jp_sales'))

	Platform	NA Sales	EU Sales	JP Sales
0	X360	226.80	117.10	3.35
1	PS3	168.26	163.10	49.39
2	PS4	108.74	141.09	15.96
3	XOne	93.12	51.59	0.34
4	3DS	82.65	61.27	100.62
5	Wii	48.56	29.60	8.97
6	WiiU	38.10	25.13	13.01
7	PC	29.88	56.70	0.00
8	DS	18.86	9.24	8.04
9	PSV	12.47	13.07	21.84
10	PSP	2.19	1.77	24.04
11	PS2	0.13	0.13	0.09

This is an exciting view into which regions gravitate towards certain platforms. I wasn't aware that X360 had so many more overall sales in North America, and I'm surprised at the distribution compared to Europe. It seems like the X360 and XOne just do not perform as well in EU and JP, as they do in NA. The same can be said about the handhelds ('3DS', 'PSV', 'PSP') and their performance in Japan vs other regions.

Cross-Regional Platform Comparison

Let's create a comparative analysis of platform performance across regions:

# Create a comparative platform analysis

# Visualize cross-regional comparison for top platforms

fig = px.histogram(regional_plotly, x='Platform', y=['NA Sales', 'EU Sales', 'JP Sales'],
                   barmode='group', height=700, labels={'value' : 'Total Sales (USD Million)', 'variable' : 'Region'},
                   title='Regional Market Analysis on Platforms')

fig.show()

#display(regional_plotly)

4.2 Regional Genre Analysis

Now let's examine genre preferences across regions:

# Function to analyze genre performance by region

def regional_genre(data: pd.DataFrame, desired_region: str) -> pd.DataFrame:
    
    selected = data[data['year_of_release'] >= 2011]
    columns = ['genre', desired_region]
    
    result = selected[columns].groupby('genre')[desired_region].sum().reset_index()
    return result.sort_values(by=desired_region, ascending=False)

regional_genre_ex = regional_genre(data, 'na_sales').merge(regional_genre(data, 'eu_sales'), on='genre', how='outer')
regional_genre_ex = regional_genre_ex.merge(regional_genre(data, 'jp_sales'), on='genre', how='outer')
regional_genre_ex = regional_genre_ex.rename(columns={'genre' : 'Genre', 'na_sales' : 'NA Sales', 'eu_sales' : 'EU Sales', 'jp_sales' : 'JP Sales'})

display(regional_genre_ex)

	Genre	NA Sales	EU Sales	JP Sales
0	Action	230.92	199.12	63.62
1	Shooter	194.43	148.02	11.92
2	Sports	108.70	88.48	11.09
3	Role-Playing	83.78	62.30	79.87
4	Misc	69.69	39.72	17.60
5	Platform	38.28	30.07	12.13
6	Racing	31.98	40.58	5.64
7	Fighting	31.63	16.24	12.38
8	Simulation	15.37	19.72	11.64
9	Adventure	14.76	13.96	12.07
10	Strategy	7.48	8.02	4.78
11	Puzzle	2.74	3.56	2.91

Cross-Regional Genre Comparison

Let's compare genre preferences across regions:

# Create a comparative genre analysis

fig = px.histogram(regional_genre_ex, x='Genre', y=['NA Sales', 'EU Sales', 'JP Sales'], barmode='group',
                   height=600, labels={'variable' : 'Region', 'value' : 'Total Sales (USD Million)'},
                   title='How do Genres Perform in each Market Region?')

fig.show()

4.3 ESRB Rating Impact Analysis

Finally, let's examine how ESRB ratings affect sales in each region:

# Function to analyze ESRB rating impact

def regional_rating(data: pd.DataFrame, desired_region: str) -> pd.DataFrame:
    
    selected = data[data['year_of_release'] >= 2011]
    columns = ['rating', desired_region]
    
    result = selected[columns].groupby('rating')[desired_region].sum().reset_index()
    return result.sort_values(by=desired_region, ascending=False)

#data.info()
#display(data)

# Analyze ESRB impact for each region

regional_rating_ex = regional_rating(data, 'na_sales').merge(regional_rating(data, 'eu_sales'), on='rating', how='outer')
regional_rating_ex = regional_rating_ex.merge(regional_rating(data, 'jp_sales'), on='rating', how='outer')
regional_rating_ex = regional_rating_ex.rename(columns={'rating' : 'Rating', 'na_sales' : 'NA Sales', 'eu_sales' : 'EU Sales', 'jp_sales' : 'JP Sales'})

display(regional_rating_ex)

fig = px.histogram(regional_rating_ex, x='Rating', y=['NA Sales', 'EU Sales', 'JP Sales'],
                   barmode='group', height=600, labels={'variable' : 'Region', 'value' : 'Total Sales (USD Million)'},
                   title='What Impact Does Rating Have on Sales?')

fig.show()

	Rating	NA Sales	EU Sales	JP Sales
0	M	308.45	250.26	26.14
1	E	170.17	141.73	36.68
2	E10+	117.72	78.00	11.23
3	T	109.31	83.16	33.39
4	EC	0.42	0.00	0.00
5	RP	0.00	0.03	0.00

*people like a Mature game*

'Grand Theft Auto V' & 'Call of Duty' really rocked the gaming market hard in favor of Action and Shooter games rated M. With Japan consuming substantially less 'M'/'Mature' content, and even consuming more titles rated 'E'/'Everyone'. This also makes sense, since a lot of handheled titles are rated 'E'/'Everyone', and Japan leads in handheld platform consumption.

Reviewer's comment v1

Well done!

Step 5 : Hypothesis Tests

—Average user ratings of the Xbox One and PC platforms are the same.

—Average user ratings for the Action and Sports genres are different.

Set the alpha threshold value yourself.

Explain:

—How you formulated the null and alternative hypotheses

—What criteria you used to test the hypotheses, and why

        #1.Average user ratings of the XOne == PC

#Null-Hypothesis: XOne['user_rating'].mean() == PC['user_rating'].mean()
#Alternate-Hypothesis: XOne['user_rating'].mean() != PC['user_rating'].mean()

xone_avg = data[(data['year_of_release'] >= 2011) & (data['platform'] == 'XOne')]['user_score'].dropna()
pc_avg = data[(data['year_of_release'] >= 2011) & (data['platform'] == 'PC')]['user_score'].dropna()

alpha = .05

results = st.ttest_ind(xone_avg, pc_avg, equal_var=False)

print(f'p-value: ', results.pvalue)

if results.pvalue < alpha:
    print(f'We reject the Null Hypothesis')
else:
    print(f'We cannot reject the Null Hypothesis')

p-value:  0.6130712247638477
We cannot reject the Null Hypothesis

*Null-Hypothesis: XOne['user_rating'].mean() == PC['user_rating'].mean()*

*Alternate-Hypothesis: XOne['user_rating'].mean() != PC['user_rating'].mean()*

With a statistical significance of 5% or .05, there is not sufficient evidence to reject the null hypothesis. We can conclude that the average user ratings of XOne and PC platforms are the same.

I chose to use 2 series objects containing the user scores of all games from both platforms respectively, and staying true to our relevant time period of 2011 - 2016. This allowed me to test the means from two statistical populations (platform) that cannot be assumed to share a statistical variance (XOne vs PC).

        #Average user ratings of Action != Sports

#Null-Hypothesis: Action['user_rating'].mean() != Sports['user_rating'].mean()
#Alternate-Hypothesis: Action['user_rating'].mean() == Sports['user_rating'].mean()

action_avg = data[(data['year_of_release'] >= 2011) & (data['genre'] == 'Action')]['user_score'].dropna()
sports_avg = data[(data['year_of_release'] >= 2011) & (data['genre'] == 'Sports')]['user_score'].dropna()

alpha = .05

results = st.ttest_ind(action_avg, sports_avg, equal_var=False)

print(f'p-value: ', results.pvalue)

if results.pvalue < alpha:
    print(f'We reject the Null Hypothesis')
else:
    print(f'We cannot reject the Null Hypothesis')

#display(action_avg, sports_avg)
#data.info()
#display(data)

p-value:  8.285648549432565e-19
We reject the Null Hypothesis

*Null-Hypothesis: Action['user_rating'].mean() != Sports['user_rating'].mean()*

*Alternate-Hypothesis: Action['user_rating'].mean() == Sports['user_rating'].mean()*

With a statistical significance of 5% or .05, there is sufficient evidence to reject the null hypothesis. We can conclude that the average user ratings of Action and Sports genres are the same.

**I chose to use 2 series objects containing the user scores of all games from both genres respectively ('Action' & 'Sports'), and staying true to our relevant time period of 2011 - 2016. This allowed me to test the means from two statistical populations (genre) that cannot be assumed to share a statistical variance ('Action' vs 'Sports').**

Reviewer's comment v1

Well done!

Step 6. Write a general conclusion

*Remarkable observations:*

1. Many titles are missing some combination of [user/critic/rating], with no observable pattern.

2. 'PC' is a veteran, emerging in 1985, and staying consistent through time into the modern era (2011 - 2016)

3. 'PC' can almost be considered the main artery in which the others branch from, and no generational successor in sight due to missing hardware 'constraints'. A constantly evolving platform, while time is a concept only to the others.

4. 'PlayStation' and 'XBox' would spear-head a surge into the gaming market, with 'Playstation' seeing 100 Million USD in sales within 4 years of launching, and 'XBox' within 8 years.

5. Our relevant time period (2011 - 2016) was selected based on observational trajectory in total number of titles being released, which inevitably impacts the rate at which developers make profit. Probable causes include developement time demand, scale of projects, etc.

6. Our data represents the 'passing of the torch', or a period in time where titles are being released on generationally different hardware ['PS3' -> 'PS4'], maybe as a way to give users the time to make the hardware change, a grace period if you will.

7. Handhelds and Consoles follow similar trends, in the interesting case of '3DS' showing an increase in sales when titles stopped releasing for the previous generation 'DS'. Probable cause is removal of hardware/software computational power difference constraints. Innovating ideas bring increase in sales. We have yet to see if the same goes for the consoles, likely.

8. 'PS3' and 'X360' struck gold with the release of 'Grand Theft Auto V' and 'Call of Duty', setting the Action & Shooter genres to be leagues ahead of others in sales.

9. Japan consumes significantly less rated 'M'/'Mature' content than the other regions [North America, Europe]. With Japan even consuming more content rated 'E'/'Everyone', in correlation to the '3DS' handheld reaching 100 Million USD in title sales, doubling their runner up 'PS3'.

10. Action and Shooter titles have gapped the other genres with similar distributions between both North America and Europe, with Role-Playing as Japan's highest selling genre.

11. 'Mature' games have a significant market lead, with 'Everyone' trailing at about 55% the same market saturation.

12. The average user ratings of [XOne/PC] and [Action/Sports] are the same.

Reviewer's comment v1

You did an excellent job analyzing the data and drawing conclusions!

It’s great that you pay attention to formulating interim conclusions in your project-this is an excellent practice that will benefit you in future projects. Keep using this approach, as it makes your analysis more structured and easier for readers and colleagues to understand.

Additionally, it’s fantastic that you included valuable insights about trends in the gaming industry. You did a great job highlighting key points and interpreting them clearly.

Overall, excellent work! Keep it up! 😊